YGENO-08730; No. of pages: 7; 4C: Genomics xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Genomics journal homepage: www.elsevier.com/locate/ygeno

Collective effects of SNPs on transgenerational inheritance in Caenorhabditis elegans and budding yeast Zuobin Zhu, Xian Man, Mengying Xia, Yimin Huang, Dejian Yuan, Shi Huang ⁎ State Key Laboratory of Medical Genetics, Central South University, 110 Xiangya Road, Changsha, Hunan 410078, PR China

a r t i c l e

i n f o

Article history: Received 25 February 2015 Accepted 6 April 2015 Available online xxxx Keywords: Collective effects Transgenerational inheritance Minor allele content (MAC) Quantitative trait loci Genetic variations SNPs RIAILs Segregants Yeasts C. elegans

a b s t r a c t We studied the collective effects of single nucleotide polymorphisms (SNPs) on transgenerational inheritance in Caenorhabditis elegans recombinant inbred advanced intercross lines (RIAILs) and yeast segregants. We divided the RIAILs and segregants into two groups of high and low minor allele content (MAC). RIAILs with higher MAC needed less generations of benzaldehyde training to gain a stable olfactory imprint and showed a greater change from normal after benzaldehyde training. Yeast segregants with higher MAC showed a more dramatic shortening of the lag phase length after ethanol exposure. The short lag phase as acquired by ethanol training was more dramatically lost after recovery in ethanol free medium for the high MAC group. We also found a preferential association between MAC and traits linked with higher number of additive QTLs. These results suggest a role for the collective effects of SNPs in transgenerational inheritance, and may help explain human variations in disease susceptibility. © 2015 Elsevier Inc. All rights reserved.

1. Introduction Understanding how naturally occurring variations in DNA sequences causes phenotypic variations in quantitative traits is a major challenge of contemporary biology. Most genetic polymorphisms such as SNPs are assumed to be nonfunctional or neutral and the field of population genetics has the neutral theory as its null hypothesis. Most complex traits are in part genetically determined and show quantitative variations among individuals within a population. Past studies have succeeded in uncovering numerous SNPs for many different traits and diseases, but the major genetic determinants or missing heritability in most cases remain to be discovered [1–8]. The neutral concept is largely built on theoretical and bioinformatic approaches and relies on a posteriori assumptions that are a priori hard to defend (e.g., using the infinite sites assumption to model finite size genomes). Recent studies have found numerous functional DNAs in the so called junk DNA regions, such as syn sites, lncRNA, pseudogene transcripts, antisense transcripts, microRNA, and transposon elements [9–11]. New theoretical frameworks based on a priori sound foundations

Abbreviations: MAF, minor allele frequency; MAC, minor allele content; HMAC, high minor allele content; LMAC, low minor allele content; QTL, quantitative trait loci; FDR, false discovery rate. ⁎ Corresponding author. E-mail address: [email protected] (S. Huang).

have also appeared in recent years that can coherently account for the reality of far more functional DNAs as well as all other major known facts of evolution and population genetics[12–16]. Most cellular components exert their functions through interactions with other components, and this network of interactions is not random but is characterized by a core set of organizing principles [17]. So the phenotypic impact of a defect is not determined solely by a mutated gene, but also by the functions of components with which the gene and its products interact [18,19]. Genome wide SNP typing experiments have made it possible to experimentally test the collective effects of multiple mutations or SNPs. While a common SNP may have little individual effect, there is a strong collective effect of SNPs on numerous traits [20,21]. Different cell types of an individual organism carry the same DNA but manifest different traits or functions due to different epigenetic programming. A cell or organism may also acquire new traits by way of epigenetic reprogramming through interaction with the environment. It is well established that both inherent traits and acquired traits can be transmitted through multiple generations with some traits more stable than others [22,23]. But the relationship between the stability of such transgenerational inheritance and the degree of genetic variations in an individual or cell has yet to be explored. Also unknown is the relationship between genetic variations and sensitivity to environmental factors. Better understanding of such relationships may help explain the well-known variations in disease susceptibility in human individuals when exposed to the same environmental pathogenic factors.

http://dx.doi.org/10.1016/j.ygeno.2015.04.002 0888-7543/© 2015 Elsevier Inc. All rights reserved.

Please cite this article as: Z. Zhu, et al., Genomics (2015), http://dx.doi.org/10.1016/j.ygeno.2015.04.002

2

Z. Zhu et al. / Genomics xxx (2015) xxx–xxx

Genetic variations are thought to be random in origins and could only be a priori destructive to any ordered processes or pathways. We here asked whether excess genetic variations can affect the transgenerational inheritance of a trait in responses to environmental factors. We made use of recombinant inbred advanced intercross lines (RIAILs) of Caenorhabditis elegans. We also used a segregant panel of the yeast Saccharomyces cerevisiae. The RIAILs are derived from C. elegans wild-type N2 (Bristol strain) and CB4856 (Hawaii strain) [24]. The F1 up to F10 progenies are intercrossed to maximize random recombination and hence allelic diversity in the offspring population, which were then randomly selected for inbreeding up to 20 generations to generate the final panel of RIAILs homozygous for almost all variants or SNPs. During the random mating and subsequent inbreeding process, there are ample opportunities for harmful variants to be negatively selected and for neutral variants to drift. Thus the frequencies of variants that exist in the established RIAILs panel are the results of both natural selection and neutral drift. We also used a panel of 124 yeast segregants derived from a cross between a laboratory strain BY4716 (BY) and a vineyard isolate RM11-1a (RM) of Saccharomyces cerevisiae. Each segregant is homozygous in nearly all SNPs and the study here used a panel of 2956 SNPs previously genotyped for these segregants [25,26]. The final panel of segregants has gone through many generations of growth during which segregants with poor ability to grow would have been selected out. Those with lethal combinations of variants would not even have been born. Thus alleles present in the final panel of segregants are a result of both selection and drift. For a given panel of RIAILs or segregants, we called minor alleles (MAs) as those parental alleles that were carried by less than half of the strains in the panel. The strains would differ in the contents of MAs that each carries, and we defined “MA contents or MAC” as the total number of MAs in an individual divided by the number of SNPs scanned. Different from MA frequency (MAF), MAC is an individual measure. The genetic differences that exist between the parental strains are due to random mutations. In a panel of RIAILs or segregants, some parental alleles would be less represented or found as MAs. This could be due to random drift. Alternatively, such alleles may be slightly deleterious to certain strains and under slightly more negative selection than positive during the growth and propagation of strains under laboratory conditions. Furthermore, if a strain is enriched with these deleterious MAs by chance, it would be expected to have properties more likely to be under slightly negative selection. Given that a trait is typically an outcome of highly ordered biochemical processes, one predicts that strains with more MAC should have lower capacity to maintain stable inheritance of a trait, if the reason for those MAs to be minor in the final RIAILs or segregant panel is because they are slightly

deleterious. Thus, a positive result from the experiments here could resolve both the issue of neutrality for most MAs as called here and the question of genetic variations on transgenerational inheritance. 2. Results 2.1. MAC affects establishment of an epigenetically transmitted trait in C. elegans The RIAILs used in this study have 1454 genotyped nuclear SNP markers span 98.6% of the physical length of the chromosomes [27]. We used these SNPs to calculate MAC for each RIAIL and found a great variation in MAC among the RIAILs (MAC from ~0.2 to ~0.7) (Supplementary Table S1). A life-long olfactory imprint is known to get established by epigenetic mechanisms during the first larval stage in C. elegans [23]. The olfactory imprint can be stably inherited through many generations if C. elegans were exposed for several generations to an environmental stimulus such as benzaldehyde (BA) [23]. Here we examined two indexes of the BA induced olfactory trait among different RIAILs: one is the number of generations required for training in the BA environment to produce a stably inherited olfactory imprint; the other the fold of changes in chemotaxis index (CI) under training condition relative to the normal condition. We divided the RIAILs into high and low MAC groups of 9 strains each (Supplementary Table S1). The data showed that high MAC (HMAC) strains needed less generations of benzaldehyde training to gain a stable imprint (Fig. 1A) and consistently showed a greater change from normal after benzaldehyde training, relative to the low MAC (LMAC) group (Fig. 1B). Motility test showed no difference in locomotor index (LI) between HMAC and LMAC strains (Fig. 1C). The results indicated a higher capacity of high MAC strains to acquire a new epigenetic trait or to lose an inherent trait. Experiments to test the stability of the newly acquired imprint are too extensive to be practically feasible due to the large numbers of strains involved. This issue may be better and more feasibly addressed by using an organism such as yeasts with much shorter generation times and more easily measurable traits as described below. 2.2. MAC affects both establishment and maintenance of an epigenetically transmitted trait in yeasts To further study the role of MAC in transgenerational inheritance, we made use of the 124 yeast segregants derived from a cross between a laboratory strain BY4716(BY) and a vineyard isolate RM11-1a(RM) of Saccharomyces cerevisiae [26,28]. We used a panel of 2956 SNPs previously genotyped for these segregants to determine the MAF of each SNP in the population of 124 segregants [20,25,26]. Of the 2956 SNPs

Fig. 1. Comparison of HMAC and LMAC group C. elegans RIAILs in transgenerational inheritance traits. (A) The number of generations needed for training in benzaldehyde (BA) for acquiring a stable olfactory imprint. (B) The fold of change in CI after vs. before BA training. (C) The locomotor index. About 200 C. elegans of each strains were used in the experiment. *P b 0.05, Student's t test. Data are mean ± SEM (standard error of the mean). Experiment was repeated 3 times; the locomotion assay was repeated 2 times.

Please cite this article as: Z. Zhu, et al., Genomics (2015), http://dx.doi.org/10.1016/j.ygeno.2015.04.002

Z. Zhu et al. / Genomics xxx (2015) xxx–xxx

scanned, 121 have MAF 0.5 and were considered non-informative. Of the remaining 2835 SNPs, 1589 MAs were from BY and 1246 from RM (P b 0.01, chi-square test, Supplementary Table S2). We then calculated the MA content (MAC) in each segregant which ranged from 0.3 to 0.6 (Supplementary Table S2). We determined whether MAC is simply a measure of the amount of parental alleles in a segregant or whether high MAC values mean more BY alleles since more BY alleles were found as MAs. A segregant could have MAs by either chance or natural selection. If by chance, it would be a high probability event for a BY allele to be present in the high MAC segregants. If by natural selection, it would not be so necessarily. If BY alleles are more deleterious than RM alleles, a segregant of high MAC value may not survive if it is enriched with BY alleles. Indeed, we found that one cannot predict parental allele content from MAC values (Supplementary Table S2). For example, segregant 17_1_a has a low MAC value 0.408 and less BY alleles than RM alleles (1381 vs. 1431). On the other hand, another segregant 15_6_c with low MAC value 0.413 has more BY than RM alleles (1584 vs. 1245). Segregant 1_1_d has the highest MAC 0.591 and yet has less BY than RM alleles (959 vs. 1336, the number does not add to 2835 because some SNPs have no genotype information). There were no enrichments of either BY or RM alleles in high MAC segregants relative to low MAC segregants (14 with more BY alleles in 20 high MAC segregants vs. 12 with more BY alleles in 20 low MAC segregants, P N 0.05, chi-square test, Supplementary Table S2). There is often a lag phase when microorganisms adapt themselves to new conditions, during which they acquire nutrients from the new growth medium and have active metabolism but not yet able to divide [29]. The length of the lag phase λ is an inherent trait of each organism. Viable count enumeration and optical density (OD) analysis are commonly used to estimate the lag phase and both are similarly useful [30]. Here we used the OD measurement to estimate the lag phase. We divided the segregants into two groups of 10 each, the high MAC (HMAC) group with MAC 0.5–0.6 and the low MAC (LMAC) group with MAC 0.3–0.4. Two lag phase traits were examined. The first is the fold of ethanol-training induced change in lag phase length (FET), which is the ratio of lag phase length after 14 days of ethanol-training over that in normal medium. We measured the lag phase length after 14 days of adaptive training in ethanol media. We selected 14 days because time course experiments showed an insignificant change in lag phase after 7 days of training while a similar change to that of 14 days after 30 days of training (Supplemental Fig. S1). At the 14th day in ethanolcontaining media, the lag phase of all strains became shorter but the decrease was more dramatic for the HMAC group, indicating a greater ethanol induced change in the HMAC group relative to the LMAC group (Fig. 2A, Supplementary Table S3). The results suggest that high MAC segregants were less able to maintain the inherent trait of lag phase length or more able to acquire a newly induced trait, consistent with the above described C. elegans results. The second trait we examined is the fold of recovery from ethanol treatment (FRE), which is the ratio of lag phase length after versus before recovery in ethanol free medium. After 8 days of culture in ethanol free medium, the HMAC group returned back to the longer lag phase trait much more dramatically than the LMAC group (Fig. 2B, Supplementary Table S3). The results indicated a poor memory of the acquired trait for the HMAC relative to the LMAC group. To confirm these results, we examined the response of segregants to sodium chloride treatment using the same experimental procedures except that ethanol was replaced by sodium chloride. Overall, the results of the sodium chloride treatment were similar to those of the ethanol experiment, although less dramatic (Supplementary Fig. S2 and Table S3). 2.3. MAC is independent of dominant effect SNPs The above association of MAC with complex traits could be due to a few major effect SNPs that happen to associate with both traits and

3

Fig. 2. Comparison of HMAC and LMAC group yeast segregants in transgenerational inheritance traits. (A) The fold of change from the inherent lag phase length to the acquired lag phase length (trained in ethanol). (B) The fold of change from the recovered lag phase to the acquired lag phase (trained in ethanol). *P b 0.05, Student's t-test. Data are means ± SEM. Each experiment was repeated 3 times.

MAC. To test this possibility, we searched for potential major effect SNPs by using the quantitative trait association option of the PLINK software [31]. We analyzed the two C. elegans traits described above, the number of generations needed for acquiring the stably transmitted imprint (generations) and the degree of change induced by the odorant. There were 17 SNPs associated with generations but not with CI at the false discovery rate (FDR) 0.05 (Supplementary Table S4). But none of these 17 SNPs showed any association with MAC (Supplementary Table S5). For the two traits described for yeast segregants, FET and FRE, we did not find any SNPs associated with these traits at FDR 5% (Supplementary Table S6). Nonetheless, we selected 11 top ranked SNPs associated with FET and 32 top ranked SNPs associated with FRE to test whether these SNPs may be associated with MAC (P b 0.01, Wald test asymptotic). One SNP 8057_i_at_x13 weakly associated with FRE was found linked with MAC (Supplementary Table S7). No SNPs weakly associated with FET were found linked with MAC at FDR 5% (Supplementary Table S7). BY allele was the minor allele of the SNP of 8057_i_at_x13, and the single SNP 8057_i_at_x13 was located in the ncRNA of snR62 on chromosome 15. Next, we determined whether the single SNP 8057_i_at_x13 could be responsible for the above observed association between MAC and lag phase traits. We found that strains carrying the minor allele (BY allele) of 8057_i_at_x13 had higher FRE (Fig. 3). High MAC strains however showed trends of higher FRE values than low MAC strains regardless of the genotypes at SNP 8057_i_at_x13. These results suggest that the effect of MAC on lag phase traits is not due to some individual SNPs that happen to associate with both MAC and lag phase traits. While one could detect certain individual SNP weakly associated with both MAC and transgenerational inheritance traits, it cannot fully account for the effect of MAC. 2.4. MAC and gene expression in yeast segregants To study whether MAC may correlate with gene expression, which may underlie its association with complex traits, we studied a yeast microarray dataset from the literature [32]. As a comparison, we also tested the collective effect of BY parental alleles or BY allele content (BAC). We found 75 genes regulated by MAC and 45 genes by BAC and 3 genes by both MAC and BAC under normal culture conditions, by using the most stringent criterion according to the SAM analysis software (Supplementary Table S8). In the presence of 1% ethanol, we found 153 genes regulated by MAC, 46 genes by BAC, and 6 genes by both MAC

Please cite this article as: Z. Zhu, et al., Genomics (2015), http://dx.doi.org/10.1016/j.ygeno.2015.04.002

4

Z. Zhu et al. / Genomics xxx (2015) xxx–xxx

Fig. 3. The role of MAC and the major effect SNP 8057_i_at_x13 in lag phase length. The SNP 8057_i_at_x13 was weakly associated with both FRE and MAC. BY is the minor allele of 8057_i_at_x13 and RM the major allele. The fold of recovery from ethanol treatment (FRE) is shown, which is the ratio of the lag phase length after 7 days of recovery relative to that after ethanol training. *P b 0.05, Student's t-test. Data are means ± SEM. Each experiment was repeated 3 times.

and BAC (Supplementary Table S9). These data indicate a role for gene expression control in MAC association with complex traits and further confirm the notion addressed above that MAC is not simply a measure of parental alleles. 2.5. MAC and the number of known additive loci of a trait That a complex trait in an individual is correlated with the total amount of SNPs, i.e., MAC, in the individual suggests a role for multiple genetic loci acting in an additive fashion. To verify this, we analyzed a published study on a large panel of yeast segregants derived from a cross between a variant strain of BY4716 and a variant of RM11-1a, which identified a large number of additive QTLs associated with each studied trait [8]. The number of additive loci ranged from 5 to 29 (average 13) for the 41 traits studied, although that study cannot possibly identify all possible QTLs for a trait due to experimental limitations such as sample size. We determined the MAs of the 392 SNPs genotyped for these segregants and calculated the MAC for each of the 1009 segregants (Supplementary Table S10). Of the MAs, 185 were BY alleles and 207 RM alleles (P N 0.05, chi-square test). Fourteen traits were significantly linked with MAC by Spearman correlation analysis, and 7 of them remain significant after multivariate regression analysis (Supplementary Table S11). The 14 traits with stronger linkage to MAC have on average 15.4 QTLs versus 12.3 for the 26 lower linkage traits (P b 0.05, Student's t test), indicating higher number of additive QTLs for traits linked with MAC (Fig. 4A). In addition, the 7 traits most definitively linked to MAC as determined by multivariate analysis have on average 18.1 additive QTLs. The results suggest that the link between MAC and the quantitative variations of a trait phenotype is due to the additive effect of a large number of QTLs as well as the functional difference between MAs and the major alleles. We also found that traits with stronger linkage to MAC have less number of pairs of interacting QTLs (Fig. 4B). Furthermore, the interacting QTLs contributed less to traits linked with higher MAC (Fig. 4C). The results suggest that excess genomic variations may weaken the interactions among genes. 3. Discussion The results here suggest that the effects of environmental factors on inherent traits may vary depending on the seemingly normal genetic variations in an organism or cell. Although it is well established that large effect mutations can affect susceptibility to environmental factors, this study may help explain the role of common SNPs in the stability of

Fig. 4. Correlation between traits and additive QTLs. (A) Number of additive QTLs for traits linked or not linked with MAC. (B) Number of pairs of interacting QTLs for traits linked or not linked with MAC. (C) Negative correlation between the number of additive QTLs and the variance explained by interacting QTLs.

transgenerational inheritance. Furthermore, the stability of a newly acquired epigenetically transmitted trait is also affected by common SNPs. Phenotypic variations in the C. elegans RIAILs or yeast segregants can be partitioned into the contribution of heritable genetic factors and measurement errors or other random environmental effects. In the experiments here, gene–environment interactions for a given trait should be absent as all the segregants or RIAILs were grown simultaneously under uniform conditions. For the progeny population derived from the cross between BY and RM, the BY alleles were significantly less represented (mostly minor alleles) in the 124 strain panel used here. On the other hand, BY alleles were slightly or non-significantly more represented in the 1009 strain panel used in the Bloom et al. study [33]. These differences could result from variations in the SNPs ascertained, the parental strains, and the growth conditions, which could all affect the selective pressure on SNPs. That these SNPs or MAs are not neutral was supported by both the absence of a correlation between BY allele content and MAC and the trait difference between HMAC and LMAC groups. These results here support the previous conclusion on the non-neutrality of most common SNPs [20,21]. The poor ability of the HMAC group of both worms and yeasts in maintaining an inherent trait is in line with the observation that most MAs are minor because they are under slightly negative selection with regard to survival of the segregants under laboratory conditions. It is expected that deleterious MAs would adversely affect some orderly biochemical/physiological pathways, which provides the most parsimonious explanation to the observed results. Also consistently, HMAC group of yeasts were less able to maintain a newly acquired epigenetic trait.

Please cite this article as: Z. Zhu, et al., Genomics (2015), http://dx.doi.org/10.1016/j.ygeno.2015.04.002

Z. Zhu et al. / Genomics xxx (2015) xxx–xxx

Difference in the stability of lag phase length exists in the parental strains and is hence presumably due to genetic or SNP differences between the strains. However, that difference alone reveals little on whether such difference is due to multiple SNPs and whether SNPs associated with low trait-stability are in general more likely to be deleterious. The results here clarify these issues. MAC linked traits tend to have higher number of identified additive QTLs. While some of these linkages have weak P values, which would be treated as false positives by Bonferroni correction, it may be more prudent here to not to use such correction since there is a high risk of false negatives [34]. But these weak associations should be verified by future studies, while the study here is more conclusive on the stronger ones as found by multivariate analysis. Overall, the effect of MAC suggests that the ethanol induced lag phase response is determined by a large number of genetic loci acting in an additive manner. Indeed, ethanol tolerance in yeast is highly heritable and thought to be determined by as many as 251 genes as well as a large number of additive QTLs [35]. If major alleles represent the favored allele to an ordered biochemical pathway, then a mutation that changes a major allele to a minor one could be regarded as a random disruption to the ordered pathway. Thus more minor alleles mean more mutations and hence more randomness or disorder in the system, which may adversely affect certain traits. Most common MAs may not have large effects individually but a group of them together over a threshold limit may have significant effects. The effect could be very minor so that the MAs would not be rare in frequency or under strong negative selection. Most genome wide association studies (GWAS) or other existing methods may not be able to detect such minor effect SNPs individually and thus create the artificial problem of “missing heritability” [2]. In reality, however, most of what is missing may be in the so called neutral SNPs, whose collective effect can now be detectable by the concept and method of MAC as shown here and elsewhere [20,21]. Our results further showed a correlation between enrichment of MA contents and mRNA expression, extending previous work on eQTLs [25]. Future work may reveal the mechanisms by which a large number of SNPs or eQTLs may affect the expression of an individual gene. Most inherent traits and acquired traits are determined by epigenetic programming. There are great variations in the transgenerational epigenetic stability of inherent and acquired traits [22,23]. The work here shows an important role of seemingly normal genetic variations in the stability of transgenerational inheritance or epigenetic inheritance, providing direct evidence for the previously proposed common sense notion of an inverse relationship between genetic diversity and epigenetic complexity or orderness [12–16]. It also has implications for disease prevention and treatment. Individuals with more SNP minor alleles may be more susceptible to environmental pathogens due to the adverse effect of MAs on inherent traits. But they may also be more easily treatable if treatment was administered relatively early before the disease has progressed past the threshold of no return, because the acquired disease trait may be less stably maintained in these individuals.

4. Materials and methods 4.1. MAF and MAC calculation The calculation was done as previously described [20,21]. The SNP datasets were obtained from R. Brem, E. Smith, and L. Kruglyak. The MAF of each SNP in the panel of RIAILs or segregants was calculated by SNP Tools for Microsoft Excel and PLINK [31,36]. From such MAF data, we obtained the MA set, which excluded non-informative SNPs with MAF = 0 or 0.5 in the panel. The MA set was equivalent to the genotype of an imagined individual who is homozygous for all the MAs. The MAC of each segregant was then determined by matching the genotype of a sergeant with the MA set; the number of identical genotypes was scored as the number of MAs for the segregant [20,21].

5

The MAC of a strain was calculated by dividing the number of MAs carried by the strain by the number of total SNPs scanned. 4.2. Strains and media C. elegans RIAILs were gifts from L. Kruglyak and yeast segregants were gifts from R. Brem. C. elegans were cultivated at 20 °C on normal growth medium (NGM) and seeded with the E. coli OP50. Yeast segregants were cultivated in the YPD media consisted of 1% yeast extract, 2% glucose, and 2% peptone. Ethanol treatment used 7% v/v ethanol in YPD media. Sodium chloride treatment used 0.8 M sodium chloride in YPD media. All growth was performed in an Orbital Shaker at 200 revolutions per minute (RPM) and 30 °C. 4.3. Benzaldehyde training assays About 100 C. elegans eggs were placed on 6 cm Petri dishes filled with 8 ml of agar medium containing E. coli OP50. 200 μl benzaldehyde (v/v; 1/400, diluted in ethanol) was presented at the lid of the dish. The dishes were sealed well with the sealing film and placed upside down at 20 °C. The F1 adults (first generation) were divided into two groups. Group one was used for continuous passage in the presence of benzaldehyde until the olfactory imprints of the C. elegans can be stably inherited for more than three generations. The other group was used for chemotaxis assays. 4.4. Chemotaxis assays Chemotaxis assays were carried out as previously described [37]. If a strain showed a high level of olfactory imprints for three generations, it was considered to have gained a stable epigenetic memory. The 9 cm Petri dishes were filled with 15 ml of agar medium (without E. coli or axenic bacteria). On one side of the agar, a 1 μl drop of benzaldehyde (diluted in ethanol) was presented, and a 1 μl drop of ethanol was presented on the other side. Prior to this test, 1 μl of 1 M NaN3 was applied to the centers of the two test spots to immobilize the animals (Fig. 5A). After a predetermined time period necessary for worms to reach from the center to the edge of the dish, which varied from strain to strain, a chemotaxis index (CI) was calculated to measure the preference toward the test odorant according to the formula ((# in A × 2 + # in B) − (# in D × 2 + # in C)) / (total # of C. elegans on the plate) (Fig. 5A). A positive CI indicates an attraction, and a negative indicates an aversion. 4.5. Locomotor assay Locomotor assay was performed as described previously [38]. A grid with four concentric circles was used (Fig. 5B). Adults were placed at the center of the grid and allowed to disperse freely. After a predetermined time period, animals were counted. The locomotor index (LI) was calculated as LI = (# in B + # in C × 2 + # in D × 3) / (total number of worms on the plate). Those animals that crawled up the side of the plate were considered lost. Using this assay, we first tested the motility of different RIAILs to determine the appropriate time required for the worms to move to the edge of the dish. This time was used for both the chemotaxis assay and the locomotor assay. 4.6. Growth curve and lag phase phenotype determination Lag phase length phenotype was determined by measuring growth curve for each segregant. The growth curve was drawn with cultivation time for x axis and ln(OD600) for y axis. The duration of the lag phase λ can be quantified as the time obtained by extrapolating the tangent at the exponential part of the growth curve, back to the inoculum level [39]. To determine growth curve, segregants were cultivated overnight to OD600 = 1, followed by transferring a portion into 6 ml of fresh media

Please cite this article as: Z. Zhu, et al., Genomics (2015), http://dx.doi.org/10.1016/j.ygeno.2015.04.002

6

Z. Zhu et al. / Genomics xxx (2015) xxx–xxx

Fig. 5. Process map of chemotaxis assays. (A) Schematic of the chemotaxis assays. Animals were put in the starting positions as shown. After a predetermined time period necessary for worms to reach from the center to the edge of the dish, the weighted CI was calculated as ((# in A × 2 + B) − (# in D × 2 + C)) / (total number of C. elegans on the plate). (B) Schematic of the locomotion assay. Worms were placed at the center of the agar plate A. After a predetermined time period, the number of worms was counted. The locomotor index (LI) was calculated as LI = (# in B + # in C × 2 + # in D × 3) / (total number of worms on the plate).

in a 15 ml round bottom centrifuge tube with OD600 adjusted to 0.03. To determine growth curve in normal media, the culture was grown for 2 days in normal media with OD600 measured every 2 h using an automatic microplate reader. To determine growth curve in ethanol containing media, the culture was grown for 72 h in media containing 7% ethanol with OD600 measured every 2 h. Each experiment measured triplicates of each segregant and the average value of triplicates was used to draw the growth curve. The experiments were repeated three times. To determine whether ethanol can induce lag phase change, segregants were either grown in normal or 7% ethanol containing media for 14 days with fresh media change every 48 h. The growth curves of the segregants were then determined in the presence of 7% ethanol as described above to determine lag phase length. To measure loss of the ethanol induced shorter lag phase trait, we trained the segregants in the YPD media containing 7% (v/v) ethanol for 14 days with fresh media change every 48 h. After 14 days of such training, the segregants were cultivated in normal YPD media for 7 days to induce loss of any ethanol induced trait with fresh media change every 24 h (the cells were washed with new media before transfer). The lag phase of each segregant in response to ethanol was then determined by performing growth curve experiment in the presence of ethanol as described above. 4.7. Statistical methods Differences in lag phase length were examined by Student's t test, two tailed. Spearman correlation used GraphPad Prism5 and InStat3. q-Value estimation for false discovery rate control was done using R package ‘qvalue’ [40]. The PLINK software package (v1.07) with the quantitative trait association option was used to search for linked SNPs [31]. Gene expression datasets for the segregants were from previous studies [32]. The correlation between MAC and gene expression was analyzed using the Significance Analysis of Microarrays (SAM) software with 2000 sample permutations. SAM uses permutations to estimate the false discovery rate (FDR) and an adjustable threshold allows for control of the FDR [41]. SAM adopts q-value as the lowest FDR at which the gene is called significant.

Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ygeno.2015.04.002. Acknowledgments We thank R. Brem, E. Smith, M. Rockman, and L. Kruglyak for research materials or technical assistance. This work was supported by the National Natural Science Foundation of China grant 81171880 and the National Basic Research Program of China grant 2011CB51001 (S.H.). References [1] D.F. Conrad, D. Pinto, R. Redon, L. Feuk, O. Gokcumen, Y. Zhang, J. Aerts, T.D. Andrews, C. Barnes, P. Campbell, T. Fitzgerald, M. Hu, C.H. Ihm, K. Kristiansson, D.G. Macarthur, J.R. Macdonald, I. Onyiah, A.W. Pang, S. Robson, K. Stirrups, A. Valsesia, K. Walter, J. Wei, C. Tyler-Smith, N.P. Carter, C. Lee, S.W. Scherer, M.E. Hurles, Origins and functional impact of copy number variation in the human genome, Nature 464 (2010) 704–712. [2] T.A. Manolio, F.S. Collins, N.J. Cox, D.B. Goldstein, L.A. Hindorff, D.J. Hunter, M.I. McCarthy, E.M. Ramos, L.R. Cardon, A. Chakravarti, J.H. Cho, A.E. Guttmacher, A. Kong, L. Kruglyak, E. Mardis, C.N. Rotimi, M. Slatkin, D. Valle, A.S. Whittemore, M. Boehnke, A.G. Clark, E.E. Eichler, G. Gibson, J.L. Haines, T.F. Mackay, S.A. McCarroll, P.M. Visscher, Finding the missing heritability of complex diseases, Nature 461 (2009) 747–753. [3] M.C. O'Donovan, N. Craddock, N. Norton, H. Williams, T. Peirce, V. Moskvina, I. Nikolov, M. Hamshere, L. Carroll, L. Georgieva, S. Dwyer, P. Holmans, J.L. Marchini, C.C. Spencer, B. Howie, H.T. Leung, A.M. Hartmann, H.J. Moller, D.W. Morris, Y. Shi, G. Feng, P. Hoffmann, P. Propping, C. Vasilescu, W. Maier, M. Rietschel, S. Zammit, J. Schumacher, E.M. Quinn, T.G. Schulze, N.M. Williams, I. Giegling, N. Iwata, M. Ikeda, A. Darvasi, S. Shifman, L. He, J. Duan, A.R. Sanders, D.F. Levinson, P.V. Gejman, S. Cichon, M.M. Nothen, M. Gill, A. Corvin, D. Rujescu, G. Kirov, M.J. Owen, N.G. Buccola, B.J. Mowry, R. Freedman, F. Amin, D.W. Black, J.M. Silverman, W.F. Byerley, C.R. Cloninger, Identification of loci associated with schizophrenia by genome-wide association and follow-up, Nat. Genet. 40 (2008) 1053–1055. [4] S.M. Purcell, N.R. Wray, J.L. Stone, P.M. Visscher, M.C. O'Donovan, P.F. Sullivan, P. Sklar, Common polygenic variation contributes to risk of schizophrenia and bipolar disorder, Nature 460 (2009) 748–752. [5] P.C. Sabeti, P. Varilly, B. Fry, J. Lohmueller, E. Hostetter, C. Cotsapas, X. Xie, E.H. Byrne, S.A. McCarroll, R. Gaudet, S.F. Schaffner, E.S. Lander, K.A. Frazer, D.G. Ballinger, D.R. Cox, D.A. Hinds, L.L. Stuve, R.A. Gibbs, J.W. Belmont, A. Boudreau, P. Hardenbol, S.M. Leal, S. Pasternak, D.A. Wheeler, T.D. Willis, F. Yu, H. Yang, C. Zeng, Y. Gao, H. Hu, W. Hu, C. Li, W. Lin, S. Liu, H. Pan, X. Tang, J. Wang, W. Wang, J. Yu, B. Zhang, Q. Zhang, H. Zhao, H. Zhao, J. Zhou, S.B. Gabriel, R. Barry, B. Blumenstiel, A. Camargo, M. Defelice, M. Faggart, M. Goyette, S. Gupta, J. Moore, H. Nguyen, R.C. Onofrio, M. Parkin, J. Roy, E. Stahl, E. Winchester, L. Ziaugra, D. Altshuler, Y. Shen,

Please cite this article as: Z. Zhu, et al., Genomics (2015), http://dx.doi.org/10.1016/j.ygeno.2015.04.002

Z. Zhu et al. / Genomics xxx (2015) xxx–xxx

[6]

[7]

[8]

[9] [10] [11] [12]

[13] [14] [15]

[16]

[17]

Z. Yao, W. Huang, X. Chu, Y. He, L. Jin, Y. Liu, Y. Shen, W. Sun, H. Wang, Y. Wang, Y. Wang, X. Xiong, L. Xu, M.M. Waye, S.K. Tsui, H. Xue, J.T. Wong, L.M. Galver, J.B. Fan, K. Gunderson, S.S. Murray, A.R. Oliphant, M.S. Chee, A. Montpetit, F. Chagnon, V. Ferretti, M. Leboeuf, J.F. Olivier, M.S. Phillips, S. Roumy, C. Sallee, A. Verner, T.J. Hudson, P.Y. Kwok, D. Cai, D.C. Koboldt, R.D. Miller, L. Pawlikowska, et al., Genome-wide detection and characterization of positive selection in human populations, Nature 449 (2007) 913–918. J. Shi, D.F. Levinson, J. Duan, A.R. Sanders, Y. Zheng, I. Pe'Er, F. Dudbridge, P.A. Holmans, A.S. Whittemore, B.J. Mowry, A. Olincy, F. Amin, C.R. Cloninger, J.M. Silverman, N.G. Buccola, W.F. Byerley, D.W. Black, R.R. Crowe, J.R. Oksenberg, D.B. Mirel, K.S. Kendler, R. Freedman, P.V. Gejman, Common variants on chromosome 6p22.1 are associated with schizophrenia, Nature 460 (2009) 753–757. H. Stefansson, R.A. Ophoff, S. Steinberg, O.A. Andreassen, S. Cichon, D. Rujescu, T. Werge, O.P. Pietilainen, O. Mors, P.B. Mortensen, E. Sigurdsson, O. Gustafsson, M. Nyegaard, A. Tuulio-Henriksson, A. Ingason, T. Hansen, J. Suvisaari, J. Lonnqvist, T. Paunio, A.D. Borglum, A. Hartmann, A. Fink-Jensen, M. Nordentoft, D. Hougaard, B. Norgaard-Pedersen, Y. Bottcher, J. Olesen, R. Breuer, H.J. Moller, I. Giegling, H.B. Rasmussen, S. Timm, M. Mattheisen, I. Bitter, J.M. Rethelyi, B.B. Magnusdottir, T. Sigmundsson, P. Olason, G. Masson, J.R. Gulcher, M. Haraldsson, R. Fossdal, T.E. Thorgeirsson, U. Thorsteinsdottir, M. Ruggeri, S. Tosato, B. Franke, E. Strengman, L.A. Kiemeney, I. Melle, S. Djurovic, L. Abramova, V. Kaleda, J. Sanjuan, R. de Frutos, E. Bramon, E. Vassos, G. Fraser, U. Ettinger, M. Picchioni, N. Walker, T. Toulopoulou, A.C. Need, D. Ge, J.L. Yoon, K.V. Shianna, N.B. Freimer, R.M. Cantor, R. Murray, A. Kong, V. Golimbet, A. Carracedo, C. Arango, J. Costas, E.G. Jonsson, L. Terenius, I. Agartz, H. Petursson, M.M. Nothen, M. Rietschel, P.M. Matthews, P. Muglia, L. Peltonen, C.D. St, D.B. Goldstein, K. Stefansson, D.A. Collier, Common variants conferring risk of schizophrenia, Nature 460 (2009) 744–747. T.M. Teslovich, K. Musunuru, A.V. Smith, A.C. Edmondson, I.M. Stylianou, M. Koseki, J.P. Pirruccello, S. Ripatti, D.I. Chasman, C.J. Willer, C.T. Johansen, S.W. Fouchier, A. Isaacs, G.M. Peloso, M. Barbalic, S.L. Ricketts, J.C. Bis, Y.S. Aulchenko, G. Thorleifsson, M.F. Feitosa, J. Chambers, M. Orho-Melander, O. Melander, T. Johnson, X. Li, X. Guo, M. Li, C.Y. Shin, G.M. Jin, K.Y. Jin, J.Y. Lee, T. Park, K. Kim, X. Sim, O.R. Twee-Hee, D.C. Croteau-Chonka, L.A. Lange, J.D. Smith, K. Song, Z.J. Hua, X. Yuan, J. Luan, C. Lamina, A. Ziegler, W. Zhang, R.Y. Zee, A.F. Wright, J.C. Witteman, J.F. Wilson, G. Willemsen, H.E. Wichmann, J.B. Whitfield, D.M. Waterworth, N.J. Wareham, G. Waeber, P. Vollenweider, B.F. Voight, V. Vitart, A.G. Uitterlinden, M. Uda, J. Tuomilehto, J.R. Thompson, T. Tanaka, I. Surakka, H.M. Stringham, T.D. Spector, N. Soranzo, J.H. Smit, J. Sinisalo, K. Silander, E.J. Sijbrands, A. Scuteri, J. Scott, D. Schlessinger, S. Sanna, V. Salomaa, J. Saharinen, C. Sabatti, A. Ruokonen, I. Rudan, L.M. Rose, R. Roberts, M. Rieder, B.M. Psaty, P.P. Pramstaller, I. Pichler, M. Perola, B.W. Penninx, N.L. Pedersen, C. Pattaro, A.N. Parker, G. Pare, B.A. Oostra, C.J. O'Donnell, M.S. Nieminen, D.A. Nickerson, G.W. Montgomery, T. Meitinger, R. McPherson, M.I. McCarthy, et al., Biological, clinical and population relevance of 95 loci for blood lipids, Nature 466 (2010) 707–713. I.K.A.A. Dunham, An integrated encyclopedia of DNA elements in the human genome, Nature 489 (2012) 57–74. J.S.A.D. Mattick, The extent of functionality in the human genome, HUGO J. 7 (2) (2013), http://dx.doi.org/10.1186/1877-6566-7-2. M. Pheasant, J.S. Mattick, Raising the estimate of functional human sequences, Genome Res. 17 (2007) 1245–1253. T. Hu, M. Long, D. Yuan, Z. Zhu, Y. Huang, S. Huang, The genetic equidistance result: misreading by the molecular clock and neutral theory and reinterpretation nearly half of a century later, Sci. China Life Sci. 56 (2013) 254–261. S. Huang, The genetic equidistance result of molecular evolution is independent of mutation rates, J. Comput. Sci. Syst. Biol. 1 (2008) 92–102. S. Huang, Histone methylation and the initiation of cancer, Cancer Epigenetics, CRC Press, New York, 2008. S. Huang, Inverse relationship between genetic diversity and epigenetic complexity, 2009. (Preprint available, Nature Precedings; bhttp://dx.doi.org/10.1038/ npre.2009.1751.2). S. Huang, The overlap feature of the genetic equidistance result, a fundamental biological phenomenon overlooked for nearly half of a century, Biol. Theory 5 (2010) 40–52. A.L. Barabasi, N. Gulbahce, J. Loscalzo, Network medicine: a network-based approach to human disease, Nat. Rev. Genet. 12 (2011) 56–68.

7

[18] D.B. Goldstein, Common genetic variation and human traits, N. Engl. J. Med. 360 (2009) 1696–1698. [19] M. Vidal, M.E. Cusick, A.L. Barabasi, Interactome networks and human disease, Cell 144 (2011) 986–998. [20] D. Yuan, Z. Zhu, X. Tan, J. Liang, C. Zeng, J. Zhang, J. Chen, L. Ma, A. Dogan, G. Brockmann, O. Goldmann, E. Medina, A.D. Rice, R.W. Moyer, X. Man, K. Yi, Y. Li, Q. Lu, Y. Huang, S. Huang, Minor alleles of common SNPs quantitatively affect traits/ diseases and are under both positive and negative selection, arXiv:1209.2911, 2012. [21] D. Yuan, Z. Zhu, X. Tan, J. Liang, C. Zeng, J. Zhang, J. Chen, L. Ma, A. Dogan, G. Brockmann, O. Goldmann, E. Medina, A.D. Rice, R.W. Moyer, X. Man, K. Yi, Y. Li, Q. Lu, Y. Huang, S. Huang, Scoring the collective effects of SNPs: association of minor alleles with complex traits in model organisms, Sci. China Life Sci. 57 (2014) 876–888. [22] E. Jablonka, G. Raz, Transgenerational epigenetic inheritance: prevalence, mechanisms, and implications for the study of heredity and evolution, Q. Rev. Biol. 84 (2009) 131–176. [23] J.J. Remy, Stable inheritance of an acquired behavior in Caenorhabditis elegans, Curr. Biol. 20 (2010) R877–R878. [24] M.V. Rockman, L. Kruglyak, Breeding designs for recombinant inbred advanced intercross lines, Genetics 179 (2008) 1069–1078. [25] R.B. Brem, G. Yvert, R. Clinton, L. Kruglyak, Genetic dissection of transcriptional regulation in budding yeast, Science 296 (2002) 752–755. [26] E.O. Perlstein, D.M. Ruderfer, D.C. Roberts, S.L. Schreiber, L. Kruglyak, Genetic basis of individual differences in the response to small-molecule drugs in yeast, Nat. Genet. 39 (2007) 496–502. [27] M.V. Rockman, L. Kruglyak, Recombinational landscape and population genomics of Caenorhabditis elegans, PLoS Genet. 5 (2009) e1000419. [28] R.B. Brem, L. Kruglyak, The landscape of genetic complexity across 5,700 gene expression traits in yeast, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 1572–1577. [29] M.D. Rolfe, C.J. Rice, S. Lucchini, C. Pin, A. Thompson, A.D. Cameron, M. Alston, M.F. Stringer, R.P. Betts, J. Baranyi, M.W. Peck, J.C. Hinton, Lag phase is a distinct growth phase that prepares bacteria for exponential growth and involves transient metal accumulation, J. Bacteriol. 194 (2012) 686–701. [30] F. Baty, J.P. Flandrois, M.L. Delignette-Muller, Modeling the lag time of Listeria monocytogenes from viable count enumeration and optical density data, Appl. Environ. Microbiol. 68 (2002) 5816–5825. [31] S. Purcell, B. Neale, K. Todd-Brown, L. Thomas, M.A. Ferreira, D. Bender, J. Maller, P. Sklar, P.I. de Bakker, M.J. Daly, P.C. Sham, PLINK: a tool set for whole-genome association and population-based linkage analyses, Am. J. Hum. Genet. 81 (2007) 559–575. [32] E.N. Smith, L. Kruglyak, Gene–environment interaction in yeast gene expression, PLoS Biol. 6 (2008) e83. [33] J.S. Bloom, I.M. Ehrenreich, W.T. Loo, T.L. Lite, L. Kruglyak, Finding the sources of missing heritability in a yeast cross, Nature 494 (2013) 234–237. [34] T.V. Perneger, What's wrong with Bonferroni adjustments, BMJ 316 (1998) 1236–1238. [35] X.H. Hu, M.H. Wang, T. Tan, J.R. Li, H. Yang, L. Leach, R.M. Zhang, Z.W. Luo, Genetic dissection of ethanol tolerance in the budding yeast Saccharomyces cerevisiae, Genetics 175 (2007) 1479–1487. [36] B. Chen, S. Wilkening, M. Drechsel, K. Hemminki, SNP_tools: a compact tool package for analysis and conversion of genotype data for MS-Excel, BMC Res. Notes 2 (2009) 214. [37] W.M. Nuttley, S. Harbinder, D. van der Kooy, Regulation of distinct attractive and aversive mechanisms mediating benzaldehyde chemotaxis in Caenorhabditis elegans, Learn. Mem. 8 (2001) 170–181. [38] D. Tsui, D. van der Kooy, Serotonin mediates a learned increase in attraction to high concentrations of benzaldehyde in aged C. elegans, Learn. Mem. 15 (2008) 844–855. [39] I.A. Swinnen, K. Bernaerts, E.J. Dens, A.H. Geeraerd, J.F. Van Impe, Predictive modelling of the microbial lag phase: a review, Int. J. Food Microbiol. 94 (2004) 137–159. [40] J.D. Storey, R. Tibshirani, Statistical significance for genomewide studies, Proc. Natl. Acad. Sci. U. S. A. 100 (2003) 9440–9445. [41] V.G. Tusher, R. Tibshirani, G. Chu, Significance analysis of microarrays applied to the ionizing radiation response, Proc. Natl. Acad. Sci. U. S. A. 98 (2001) 5116–5121.

Please cite this article as: Z. Zhu, et al., Genomics (2015), http://dx.doi.org/10.1016/j.ygeno.2015.04.002